- Zuckerberg recalls telling Wang at the end of 2025 "we're going to write this check for, like, I don't know, like $100 billion, $200 billion" (53:21).
- He says the Prometheus cluster in Ohio is scaling to "larger than a gigawatt" (49:45) and the Hyperion cluster in Louisiana to 5 gigawatts (49:53).
- He puts operating a gigawatt for a year "on the order of $15 billion" (55:31) and the up-front capital cost at "probably, you know, $50 or $60 billion" (55:42).
Cheeky Pint published an interview with Mark Zuckerberg and Alexandr Wang on 12 October 2026, titled "Muse Wonder: Zuckerberg and Wang's $100B AI". The episode page says the two "join John for a pint to discuss the launch of Muse Wonder and the future of personal AI agents." Meta's investor relations page lists Zuckerberg as "Founder, Chairman and Chief Executive Officer", and Meta's page for Wang is titled "Alexandr Wang, Chief AI Officer".
Zuckerberg described a conversation at the end of 2025, saying "we're going to write this check for, like, I don't know, like $100 billion, $200 billion. I'm like, Alex, are you sure this is going to work?" (53:21). The episode page says the two discuss "what it really feels like to make a $100B+ CapEx commitment before knowing if the model will work."
On the compute itself, Zuckerberg said "we built this Prometheus cluster, we call it, in Ohio." and "That's scaling to larger than a gigawatt." (49:45). He said "And now we're building this Hyperion cluster in Louisiana that's going to scale to 5 gigawatts." (49:53). On cost, he said "to operate a gigawatt a year, call it on the order of $15 billion." (55:31) and "If you're talking about building the CapEx, it's probably closer to, you know, I mean, you pay for it upfront, so it's probably, you know, $50 or $60 billion." (55:42).
On the risk, Zuckerberg said "if AI works overall, plenty of other people want all our compute." (58:05) and "in order for it not to work out, it's not just that we have to fail, AI overall has to fail." (58:17).
On Muse Wonder in the Muse app, Zuckerberg said "because it's a lot bigger and it's more expensive to serve, we're going to start off primarily in paid subscriptions." (22:33). Wang said "as Mark mentioned, it is a much larger model. And so we're working through how to serve that most effectively to as many people as possible." (23:53).
What it means for you Our view
These are Zuckerberg's own loose figures in an interview, not a budget Meta has published, and he gives the first of them as "like, I don't know". His view of the risk is that the compute only goes unused if AI overall fails, because he expects others to want it. He also says Muse Wonder is bigger and more expensive to serve, so it will start primarily in paid subscriptions. If you plan around Meta's models, treat the amounts as his estimates and look for firm figures in Meta's own financial reporting, and ask what a paid-first start means for access to Muse Wonder on your plan.
TAGSCheeky PintDatacentresMetaMuse
